Journal of Biogeography (J. Biogeogr.) (2016) 43, 410–422

ORIGINAL Biogeography and colonization history ARTICLE of plethodontid from the Interior Highlands of eastern North America Samuel D. Martin1*, Donald B. Shepard2, Michael A. Steffen1, John G. Phillips1 and Ronald M. Bonett1

1Department of Biological Science, University ABSTRACT of Tulsa, 800 S. Tucker Drive, Tulsa, OK Aim The Interior Highlands (Ouachita Mountains and Ozark Plateau) are 74104, USA, 2Department of Biology, major physiographical regions of eastern North America and harbour many University of Central Arkansas, 201 Donaghey Ave., LSC 180, Conway, AR 72035, endemic species. Despite their close proximity, the Ozarks and Ouachitas have USA different geological histories and relatively distinct species pools. Few studies have tested the biogeographical origins of this region’s fauna, and most researchers have treated the Interior Highlands as a single unit. Here, we inferred the sources and timing of colonization of the Ozarks and the Ouachi- tas by analysing the biogeography of three genera of plethodontid salamanders (Eurycea, Plethodon and Desmognathus).

Location Eastern North America. Methods We constructed a well-sampled, time-calibrated phylogeny for the family using three mitochondrial and three nuclear genes in beast. Genetic data were primarily taken from GenBank, although we also pro- duced 76 novel sequences. Using lagrange, we reconstructed ancestral areas for North American plethodontids. We compared the frequency and timing of disper- sal events between the Ozarks and Ouachitas to other putative sources such as the Eastern Highlands (Appalachian Mountains and associated limestone plateaus). Results We inferred nine dispersal events from the Eastern Highlands to the Interior Highlands, and just two dispersal events between the geographically proximate Ozarks and Ouachitas. Following one dispersal in the Oligocene, other inter-highland dispersal events occurred in the Miocene and Pliocene, including two periods of near-synchronous movements.

Main conclusions Given the relatively limited faunal exchange between the Ozarks and Ouachitas, we conclude that either the river valley separating the Ozarks and Ouachitas is a more formidable barrier to plethodontid dispersal than barriers separating the Interior Highlands from the Eastern Highlands, or ecological/community contingencies have limited dispersal within the Interior Highlands. In our study, geographical proximity of upland islands does not correspond with faunal similarity.

*Correspondence: Samuel D. Martin, Keywords Department of Biological Science, University of Amphibia, ancestral area reconstruction, Appalachian Mountains, comparative Tulsa, 800 S. Tucker Drive, Tulsa, OK 74104, biogeography, dispersal, Mississippi River, Ouachita Mountains, Ozark Pla- USA. E-mail: [email protected] teau, Plethodontidae, sky island

(MacArthur & Wilson, 1963, 1967). The ages and connectiv- INTRODUCTION ity of ‘sky island’ habitats (Heald, 1951) have influenced the Dispersal among habitat islands generally becomes more dif- of many organisms that are unable to disperse ficult with increasing isolation and decreasing island size across lower elevations (reviewed by McCormack et al.,

410 http://wileyonlinelibrary.com/journal/jbi ª 2015 John Wiley & Sons Ltd doi:10.1111/jbi.12625 Dispersal by salamanders among North American highland areas

2009). Therefore, highland regions in geographical proximity and range continuity. However, in recent decades extensive should have more similar faunas (or more frequent gene DNA sequence data sets have been generated for eastern flow), while more distant regions should have less faunal North American plethodontids, as well as methods for diver- exchange. gence time estimation and ancestral area reconstruction The highland regions of eastern North America are major which can be used to better test these hypotheses. centres of endemism and diversity for many mesic-adapted Although the biogeography of eastern North American organisms, including (Duellman & Sweet, 1999), plethodontids has been analysed (Kozak et al., 2009; Bonett crayfishes (Crandall & Buhay, 2008) and freshwater fishes et al., 2014), previous authors treated the Ozarks and Oua- (Mayden, 1988). This is due to many factors, including their chitas as a single biogeographical unit (Interior Highlands), geological and climatic stability, complex topography and despite their distinct faunas. To date, no study has tested the drainage systems, abundant springs (limestone geology), and unique biogeographical origins of Ozark and Ouachita relatively consistent precipitation (Cross et al., 1986). The lar- plethodontid salamanders. Here, we construct a time-cali- gest highland region in eastern North America is the Eastern brated phylogenetic hypothesis for the family Plethodontidae Highlands, which includes the Appalachian Mountains and to infer the evolution of geographical ranges. From these adjacent uplands. reconstructions, we determine the geographical origins of Across the Mississippi River Valley to the west lie two Ozark and Ouachita plethodontids and compare the fre- smaller highland regions (Ouachita Mountains and Ozark quency of dispersal events among the three highland regions Plateau), referred to collectively as the Interior Highlands. of eastern North America. Lastly, we use divergence time The Ozark Plateau is a dissected limestone uplift, and the estimates to test for asynchrony of reconstructed dispersal Ouachita Mountains are a series of folded mountains, similar events. to the Appalachians but oriented on an east–west axis. Cur- rently separated by the Arkansas River Valley, these adjacent MATERIALS AND METHODS highlands were formed by disparate geological processes, but both date to the Ouachita orogeny during the Pennsylvanian Phylogenetic reconstruction (Thornbury, 1965), and have since weathered extensively. Therefore, these highlands pre-date the origin of plethodon- We sampled most recognized species of North American tid salamanders by at least 200 Myr (Mueller, 2006; Zhang & plethodontids, and several divergent lineages that potentially Wake, 2009; Bonett et al., 2014). Despite their geographical represent undescribed species, for a total of 230 terminal proximity (as close as ~38 km), the Ozarks and Ouachitas plethodontid taxa. We included two outgroup taxa, Rhya- have distinct faunas, with over 160 and 48 endemic species cotriton variegatus (Rhyacotritonidae) and Amphiuma respectively (Ouachita Ecoregional Assessment Team, The tridactylum (Amphiumidae; see Appendix S1 in Supporting Nature Conservancy, 2003; Ozarks Ecoregional Assessment Information for species authorities). These represent families Team, The Nature Conservancy, 2003). reconstructed as closely related to plethodontids in recent Few taxonomically broad studies concerning the origins of higher level molecular phylogenies of salamanders (Chippin- Interior Highland taxa have been conducted, although the dale et al., 2004; Wiens et al., 2006; Roelants et al., 2007; fish fauna was the subject of several influential (but primarily Zhang & Wake, 2009; Pyron & Wiens, 2011). We only sam- descriptive) works nearly three decades ago (Mayden, 1985, pled two to three representatives from each of the diverse 1987, 1988; Wiley & Mayden, 1985); Mayden (1985) also Central and South American plethodontid genera (tribe Boli- briefly discussed salamanders and crayfishes. Collectively, toglossini) because these in situ radiations would not impact these authors concluded that the Appalachians, Ozarks and our North American ancestral area reconstructions, and Ouachitas were contiguous until the Pleistocene, when they omitting them increased computational efficiency. We refer- were fragmented by glaciation and associated processes. In enced several published and unpublished intraspecific phylo- contrast, Crandall & Templeton (1999) found evidence for genies to inform our taxon sampling, and included multiple multiple episodes of dispersal and vicariance among highland representatives of species with particular relevance to our regions in the diverse crayfish genus Orconectes. reconstruction, especially species with geographically diver- In one of the first discussions of Interior Highland bio- gent lineages distributed in multiple regions. geography, Dowling (1956) noted that salamanders had the Our alignment included three mitochondrial genes: cyto- most informative distributions for reconstructing past bio- chrome b (Cytb), and NADH dehydrogenase subunits 2 geographical events, due to their strict habitat requirements (ND2) and 4 (ND4); and three nuclear genes: the recombina- and low dispersal abilities. This is especially pronounced in tion activating protein 1 gene (Rag1), pro-opiomelanocortin lungless salamanders of the family Plethodontidae. Dowling (POMC) and brain derived neurotrophic factor (BDNF). concluded that there were four primary colonizing elements, Most of the DNA sequences used were obtained from including what he considered a ‘pre-glacial’ group from the GenBank or other authors (716), but we also generated 74 Appalachians. As a rough metric to estimate timing of diver- novel sequences for this study following methods described gence, Dowling (1956) used taxonomic level (i.e. species ver- previously (Shepard & Burbrink, 2008; Bonett et al., 2014). sus subspecies), morphological and ecological differentiation, Newly generated sequences were deposited on GenBank (see

Journal of Biogeography 43, 410–422 411 ª 2015 John Wiley & Sons Ltd S. D. Martin et al.

Appendix S1 for sequence collection methods and GenBank [95% highest posterior density (HPD)]. Date estimates for accession numbers). crown-group plethodontids range from 99 to 65 Ma (Muel- Alignments were constructed separately for each gene ler, 2006; Roelants et al., 2007; Vieites et al., 2007; Zhang & using ClustalW in mega 5.03 (Tamura et al., 2007), and Wake, 2009; Bonett et al., 2014), but the most common esti- adjusted manually. Alignments were trimmed to regions con- mates are in the range of 65–85 Ma. Our root age prior taining at least 10 overlapping sequences. The alignment spanned most of that timeframe. Although absolute diver- matrix totalled 5520 nucleotide positions, was 56.6% com- gence dates may be somewhat inaccurate, we were primarily plete (including gaps), and contained the following numbers interested in whether relative dates overlapped, and thus of sequences per gene: Cytb (192 taxa), ND2 (129 taxa), ND4 whether any dispersal events may have been synchronous (192 taxa), Rag1 (205 taxa), POMC (73 taxa) and BDNF (39 (i.e. overlapping 95% HPDs). taxa). There was a significant amount of missing data, espe- We performed five independent MCMC searches of cially for POMC and BDNF; however, these absences gener- 1.0 9 108 generations in beast using the CIPRES Science ally do not appear to have a major impact on phylogenetic Gateway (Miller et al., 2010), logging trees and parameters reconstructions (Wiens, 2006; Wiens & Morrill, 2011) or every 10,000 generations. We applied unlinked substitution divergence dates (Zheng & Wiens, 2015), and incompletely models and clock models, linked tree models (functionally sampled genes can still improve phylogenetic resolution concatenated), a Yule process speciation prior, and uncorre- (Jiang et al., 2014). The sequences were distributed across lated lognormal relaxed clock rate priors (mean = 0, taxa (most genera had at least one sequence for all genes) SD = 1). We used logcombiner 1.7.4 to pool log and tree and the less well-sampled genes contained phylogenetic signal files from all five runs, conservatively discarding the first that helped to resolve deeper nodes. Phylogenetic analyses 25% of generations per run as burn-in and resampling every excluding POMC and BDNF failed to approach stationarity tenth logged generation to reduce file sizes. We used Tracer after twice as many generations and although the resulting 1.5 (Drummond & Rambaut, 2007) to confirm that indepen- trees recovered very similar relationships among terminal dent runs had converged and effective sample size values taxa, deeper nodes were poorly resolved (not shown). There- were > 200 for all parameters in the combined log file. We fore, we concluded the potential negative effects of missing calculated the maximum clade credibility (MCC) tree using data on topology, support values, and branch-length estima- treeannotator 1.7.4 with the combined posterior distribu- tion (Lemmon et al., 2009) were preferable to the loss of res- tion of trees. olution that occurred when these genes were excluded. Taxa without data for a given gene were included in the align- Ancestral area reconstruction ments as missing data for every nucleotide position. Our sampling strategy was designed to maximize taxon sampling After pruning two outgroups (Amphiuma and Rhyacotriton) for ancestral area reconstruction while minimizing the cre- and three Eurasian plethodontids (Karsenia koreana, Hydro- ation of chimeric taxa (samples consisting of sequences mantes genei and Hydromantes italicus) from the BEAST assembled from multiple individuals of a species), which are chronogram, we reconstructed ancestral areas for North undesirable because different genes in chimeric taxa may American Plethodontidae using the dispersal-extinction- yield discordant gene trees (Maddison, 1997). Our phyloge- cladogenesis (DEC) model in lagrange v. 20120508 (Ree & netic estimate is highly congruent topologically with prior Smith, 2008). Taxa were coded as occurring in one or more studies (see Results), indicating our sampling strategy had a of five regions (Fig. 1): (A) Ouachita Mountains; (B) Ozark negligible influence on our results. Plateau; (C) Coastal Plain plus Edwards Plateau (combined We used jModeltest 2.1.4 (Guindon & Gascuel, 2003; given that Edwards Plateau taxa are nested within Coastal Darriba et al., 2012) to determine the best substitution Plain clades); (D) Eastern Highlands (including the Appala- model for each gene. Searches among substitution models chian Mountains and northward to the Great Lakes; species were limited to models available in the beauti configurator, dispersal from the Eastern Highlands north to the Great which was used to generate the .xml file for beast 1.7.4 Lakes occurred recently (< 21 ka; Highton & Webster, 1976); (Drummond et al., 2012). To reduce computation time, we and (E) Western North America (WNA) plus Central and constrained the monophyly of the family Plethodontidae and South America (because the clade in Central and South the genera Aneides, , Desmognathus, Eurycea and America is sister to a clade in western North America). Spe- Plethodon, all of which have been supported by previous cies were coded as inhabiting an area if they occurred there, studies (Chippindale et al., 2004; Vieites et al., 2007, 2011; regardless of their extent of occurrence in that region, and Kozak et al., 2009). allowed to inhabit up to four regions (a realistic assumption To examine dispersal timing, we constrained the crown- based on the distributions of extant taxa). Transition proba- group age of Plethodontidae to 73 Æ 6 Ma (following Bonett bilities were equally weighted, and direct dispersal between et al., 2014; see references therein), which provided median WNA and the Eastern Highlands was disallowed, given that estimates for the ages of nodes reconstructed to represent salamanders would have to disperse from east to west via dispersal events (see Ancestral area reconstruction, below), as one of the intervening regions (Coastal Plain, Ozarks, or well as upper and lower confidence limits on age estimates Ouachitas). An additional lagrange analysis without this

412 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd Dispersal by salamanders among North American highland areas

Figure 1 Approximate distribution of North American plethodontid salamanders with colours corresponding to five areas used for ancestral area reconstruction coding: Orange (Ouachita Mountains), Pink (Ozark Plateau), Yellow (Coastal Plain plus Edwards Plateau), Green (Eastern Highlands, including the Appalachian Mountains northward to the Great Lakes) and Blue (western North America plus Middle and South America). The courses of the Mississippi and Arkansas rivers are represented by thick and thin dark blue lines respectively. Inset relief map of eastern North America shows the geographical separation among the Eastern Highlands, Ouachitas and Ozarks. dispersal constraint produced nearly identical results, with Plethodon also likely inhabited WNA, while the MRCA of the exception of a few basal nodes (see Appendix S2 for this eastern Plethodon was reconstructed to have inhabited either alternative reconstruction). We considered each event where the Coastal Plain (support 0.268; Fig. 2) or the Ouachitas a descendant node was reconstructed to inhabit a region (support 0.255). The MRCAs of the tribe Spelerpini and uninhabited by its parent node to be a dispersal event. genus Desmognathus inhabited the Eastern Highlands (sup- port 0.44 and 0.84 respectively). Within eastern North America, our analysis indicated fre- RESULTS quent movement from east to west, with nine reconstructed dispersal events from the Eastern Highlands and Coastal Phylogeny Plain to the Interior Highlands (Ozarks and Ouachitas). In Our phylogeny showed strong support (≥ 95% Bayesian pos- contrast, movement between the two Interior Highland terior probability) for many recognized clades and all genera regions has been infrequent, with only two reconstructed dis- (see Appendix S3 for complete beast phylogram with sup- persal events from the Ouachitas to the Ozarks (and none port values), and its topology agrees well with recent phylo- from Ozarks to Ouachitas). However, we note that two dis- genetic hypotheses for the Plethodontidae (Chippindale persal events from the Eastern Highlands were reconstructed et al., 2004; Mueller et al., 2004; Kozak et al., 2009; Pyron & as simultaneous colonizations of the Ozarks and Ouachitas; Wiens, 2011; Vieites et al., 2011; Bonett et al., 2014). Hemi- alternatively, these taxa may have colonized one Interior dactylium was recovered as sister to Bolitoglossini, in agree- Highland region first and subsequently dispersed to the ment with prior studies (Mueller et al., 2004; Vieites et al., other, potentially increasing the number of dispersal events 2011), although some studies have recovered Hemidactylium between the Ouachitas and Ozarks to four. Our alternative as sister to Bolitoglossini plus Spelerpini (Kozak et al., 2009; LAGRANGE analysis allowing direct transitions between Pyron & Wiens, 2011). The placement of Hemidactylium WNA and the Eastern Highlands (see Appendix S2) would only affect ancestral area reconstructions for the deep- indicated this may have occurred in the Interior Highlands est nodes in the tree, whereas we are primarily focused on Eurycea clade. We inferred no dispersal events from the Inte- events that occurred at more shallow nodes (e.g. within rior Highlands to the Eastern Highlands within eastern genera). We also note that Plethodon sequoyah was nested Plethodon, although Plethodon serratus likely dispersed south within P. albagula, a relationship which has been reported from the Ouachitas into the Coastal Plain. previously (Kozak et al., 2006; Wiens et al., 2006). The ancestor of Interior Highlands Eurycea was recon- structed to have colonized the Ozarks and Ouachitas simul- taneously (support 0.230) from the Eastern Highlands, Historical biogeography approximately 28.9 Ma (31.3–26.5 Ma; Fig. 3). Support was Our lagrange analysis indicated that the most recent com- lower for the ancestor initially colonizing only the Ozarks mon ancestor (MRCA) of Plethodontidae inhabited western (0.211) or Ouachitas (0.153). Subsequent vicariance and North America (WNA), or WNA plus the Eastern Highlands in situ speciation produced two Ozark endemic species and Coastal Plain (Fig. 2). Likewise, the MRCA of the genus (E. tynerensis and E. spelaea) and two Ouachita endemics

Journal of Biogeography 43, 410–422 413 ª 2015 John Wiley & Sons Ltd S. D. Martin et al.

(E. multiplicata and E. subfluvicola). A population of E. mul- support for the ancestor only inhabiting the Ouachitas and tiplicata also dispersed from the Ouachitas to the Ozarks Eastern Highlands (0.257) or the Coastal Plain and Eastern more recently (3.3 Ma, 3.6–3.0 Ma). Eurycea lucifuga Highlands (0.228). The populations in the Ouachitas were colonized the Ozarks from the Eastern Highlands in the Plio- later isolated, giving rise to P. kiamichi. Finally, P. albagula cene (4.0 Ma, 4.3–3.7 Ma), although this reconstruction was colonized the Ouachitas from the Coastal Plain/Edwards weakly supported (0.326) over colonization of the Eastern Plateau (3.5 Ma, 3.8–3.2 Ma). Our reconstructions indicated Highlands from the Ozarks (0.321). We feel the former a population-level range expansion from the Ouachitas to reconstruction is more likely given the greater genetic diver- the Ozarks (support 0.749; 3.0 Ma, 3.3–2.8 Ma), which cap- sity of E. lucifuga in the Eastern Highlands (Timpe, 2009). tured some of the ancestral variation of the parent popula- The ancestor of the ‘long-tailed’ clade (E. longicauda and tion before vicariance between the Ozarks and Ouachitas. E. guttolineata) expanded from the Coastal Plain into the Alternatively, there was weak support (0.242) for two inde- Eastern Highlands and Ozarks (10.1 Ma, 10.9–9.3 Ma). pendent dispersal events by P. albagula from the Ouachitas Although there was low support for the reconstruction at the to the Ozarks. Finer scale phylogeographical sampling is nec- parent node of the ‘long-tailed’ clade plus E. lucifuga (0.326; essary to distinguish between these two scenarios, and to Fig. 2), alternative reconstructions also supported a assess the validity of P. sequoyah. widespread ancestor for the ‘long-tailed’ clade. Subsequent The first colonization of the Interior Highlands, by the vicariant events resulted in E. longicauda in the Eastern ancestor of Interior Highlands Eurycea in the Oligocene Highlands and E. l. melanopleura in the Ozarks. (Fig. 3), far preceded other reconstructed dispersal events. The Ouachita endemic Desmognathus brimleyorum was The next colonization, by the ancestor of Desmognathus phylogenetically nested within a large clade of primarily East- brimleyorum, occurred in the Miocene several million years ern Highland species. Therefore, we are confident that the before three roughly synchronous (13.2–8 Ma) dispersal ancestor of D. brimleyorum colonized the Ouachitas from the events by ancestors of the ‘Ouachita slimy’ clade of Pletho- Eastern Highlands (support 0.624), although a different don, the ‘long-tailed’ Eurycea clade, and P. angusticlavius. topology might impact the age of the node and time of The common ancestor of P. mississippi and P. kiamichi dispersal (16.0 Ma, 17.4–14.7 Ma). dispersed out of the Eastern Highlands around the Miocene- Salamanders of the genus Plethodon have experienced Pliocene boundary. The remaining five dispersal events many more dispersal/extinction events than Eurycea and (P. serratus, E. lucifuga, P. albagula to the Ouachitas, E. mul- Desmognathus. Plethodon serratus was reconstructed to have tiplicata and P. albagula to the Ozarks) had overlapping date colonized both Interior Highland areas simultaneously (sup- estimates during the Pliocene (4.7–2.8 Ma). port 0.524) from the Eastern Highlands (support 0.289; Our alternative reconstruction, which allowed direct tran- 4.3 Ma, 4.7–4.0 Ma), and subsequently dispersed into the sitions between WNA and the Eastern Highlands, gave very Coastal Plain from the Ouachitas. Plethodon angusticlavius similar results (see Appendix S2). Most of the differences colonized the Ozarks from the Eastern Highlands (8.7 Ma, were expected; for example, several branches formerly 9.4–8.0 Ma). According to the most probable reconstruction, reconstructed as Coastal Plain (because they were transi- the ‘Ouachita slimy’ clade (P. ouachitae, P. caddoensis, and tions from WNA to Eastern Highlands) were supported as P. fourchensis) originated from vicariance of a wide-ranging direct transitions to the Eastern Highlands (MRCAs of Plethodon between the Eastern Highlands and the Ouachitas, Hemidactylium, Aneides aeneus, Desmognathus plus Phaeog- followed by vicariant speciation among different Ouachita nathus and eastern Plethodon). However, there were some mountaintops (Shepard & Burbrink, 2008, 2009, 2011). This other differences relevant to the Interior Highlands. The wide-ranging Plethodon ancestor had previously colonized MRCA of the Interior Highlands Eurycea clade was recon- the Ouachitas from the Eastern Highlands in the Miocene structed to have colonized the Ozarks only, followed by (12.2 Ma, 13.2–11.2 Ma). However, this reconstruction was colonization of the Ouachitas by the MRCA of E. multipli- weakly supported (0.505) over dispersal to the Ouachitas cata and E. subfluvicola. The MRCA of the ‘Ouachita slimy’ from the Eastern Highlands by the MRCA of the ‘Ouachita Plethodon was reconstructed to have dispersed from the slimy’ clade (support 0.480; 11.5 Ma, 12.4–10.5 Ma). The Eastern Highlands, rather than resulting from vicariance of common ancestor of P. mississippi and P. kiamichi expanded a widespread ancestral Plethodon. Finally, the MRCA of its range from the Eastern Highlands into the Coastal Plain Plethodon albagula was reconstructed as inhabiting both the and Ouachitas (support 0.341) at the Miocene-Pliocene Ouachitas and Coastal Plain, with subsequent vicariance boundary (5.4 Ma, 5.8–4.9 Ma); there was also marginal between the two areas.

Figure 2 beast chronogram of North American plethodontid salamanders, presented in two parts labelled (a and b), with five colours indicating highest probability ancestral area reconstruction in lagrange: (A) orange (Ouachita Mountains), (B) pink (Ozark Plateau), (C) yellow (Coastal Plain plus Edwards Plateau), (D) green (Eastern Highlands, including the Appalachian mountains northward to the Great Lakes), and (E) blue (western North America plus Central and South America). Other colours represent area combinations: grey (3+ areas, labelled), red (A+B), brown (A+D), purple (B+D), light green (C+D) and dark teal (C+E). LAGRANGE support values < 75% labelled; all unlabelled nodes received at least 75% support for ancestral area reconstruction.

414 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd Dispersal by salamanders among North American highland areas

Journal of Biogeography 43, 410–422 415 ª 2015 John Wiley & Sons Ltd S. D. Martin et al.

Figure 2 Continued.

416 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd Dispersal by salamanders among North American highland areas

likelihood of colonization with abundant and wide-ranging species being more likely colonizers. Little is known about population sizes of most plethodontids and species distribu- tions are highly variable in size. Genetic data and distribu- tion models for some extant species indicate that population sizes and geographical ranges have expanded and contracted historically (Shepard & Burbrink, 2008, 2009, 2011; Pelletier et al., 2014). However, these traits would be difficult or impossible to determine for the ancestral lineages that were inferred to have colonized new regions. Despite the ancient age of the Interior Highlands (> 300 Ma) and plethodontids (~73 Ma), only one extant, endemic lineage of plethodontids arrived during the Paleogene (Interior Highlands Eurycea; Fig. 3). This lineage subsequently diversified in situ within both the Ozarks and the Ouachitas, including considerable diversification in life history and mor- phology (Bonett & Chippindale, 2004). The ancestor of the ‘Ouachita slimy’ clade (Plethodon ouachitae group) colonized the Ouachitas in the Miocene and diversified into several eco- logically similar, parapatric and allopatric species (Shepard & Burbrink, 2008, 2009, 2011). Other Miocene colonizers of the Interior Highlands showed no evidence of in situ diversifica- tion (ancestors of Desmognathus brimleyorum, Eurycea longi- cauda melanopleura, and P. angusticlavius). The remaining plethodontid lineages colonized the Interior Highlands more recently, in the Pliocene. While Dowling (1956) concluded that the troglobitic ‘Typhlotriton’(Eurycea spelaea) was the only Interior Highland faunal element older than 2.5 Ma, we found that all plethodontid colonizations were considerably Figure 3 Estimated node ages (from beast chronogram) of older than the Pleistocene. Both Nothonotus darters and lagrange dispersal events (according to reconstruction) by amblyopsid cavefishes probably dispersed to the Interior plethodontid salamanders to one or both of the Interior Highlands in the Miocene as well as more recently, in the late Highland regions (Ozark Plateau or Ouachita Mountains). Error bars represent upper and lower age estimates (95% Highest Pliocene/early Pleistocene (Near & Keck, 2005; Niemiller Posterior Density) based on a crown-group age of 73 Æ 6Ma et al., 2013). for Plethodontidae (following Bonett et al., 2014). Nodes are One possible explanation for the apparent rarity of ‘old’ arranged from oldest to youngest. Black diamonds represent dispersal events (e.g. Oligocene) was the smaller pool of dispersal events from the Eastern Highlands or Coastal Plain to potential colonizers at that time. According to our recon- the Interior Highlands, and grey diamonds represent dispersal structions, only six plethodontid lineages were present in the events between the Ozarks and Ouachitas. Eastern Highlands when the MRCA of Interior Highlands Eurycea dispersed (31.3–26.5 Ma; Fig. 3). Other lineages may have existed, but gone extinct without leaving extant descen- DISCUSSION dants. However, by the time the ancestor of D. brimleyorum dispersed to the Interior Highlands in the Miocene there Colonization of the Interior Highlands were already nearly 20 plethodontid lineages present in the The Eastern Highlands are the largest upland region in east- Eastern Highlands, with over 70 by the end of the Miocene ern North America, and historically contained the largest (Fig. 2). This growing species pool may have provided more pool of species for potential colonizations of the Ozarks and potential species for colonization of the Interior Highlands. Ouachitas (Kozak & Wiens, 2012). Our biogeographical In contrast to the frequency of colonizations from the East- reconstructions supported the Eastern Highlands as the ori- ern to the Interior Highlands, there were few dispersal events gin for Interior Highlands Desmognathus, Eurycea, and between the Ouachitas and Ozarks. Plethodon (Fig. 2). The Eastern Highlands are also thought to be the source for other Interior Highland fauna, including Infrequent dispersal within the Interior Highlands Nothonotus darters (Near & Keck, 2005), Orconectes cray- fishes (Crandall & Templeton, 1999; and references therein), Dispersal among habitat islands should be more common with and numerous insects (Allen, 1990). Population and decreasing geographical separation (MacArthur & Wilson, geographical range sizes of species may also influence the 1963, 1967); therefore, we expected upland regions in close

Journal of Biogeography 43, 410–422 417 ª 2015 John Wiley & Sons Ltd S. D. Martin et al. proximity to have more frequent faunal exchange. Mayden and/or subterranean environments, such as Eurycea (Bonett (1987) concluded that the fish faunas of the Ozarks and & Chippindale, 2004; Bonett et al., 2014). Ouachitas were more similar to one another than to the In contrast, the Ouachitas are composed of several narrow, Eastern Highlands; however, we found the opposite pattern in east–west trending mountain ranges with short, steep drai- plethodontid salamanders. Few plethodontids have dispersed nages, scattered across low-relief lowlands. Surface stream between the Ozarks and Ouachitas despite their geographical flow is more intermittent and the Ouachitas support only proximity (as close as 38 km), while the more distant Eastern two Eurycea. There are at least seven species of Plethodon in Highlands have been the source for multiple colonizations of the Ouachitas, compared to three in the Ozarks. These the Interior Highlands. Our findings are consistent with a direct-developing, terrestrial salamanders are less dependent biogeographical reconstruction for Nothonotus darters indicat- on standing water sources compared to species with aquatic ing rare dispersal among highland regions of eastern North larvae, but still require mesic woodland microhabitats. The America (Near & Keck, 2005), a non-sister relationship higher frequency of inter-highland movements by Plethodon between the Ozarks and Ouachitas in the Notropis rubellus may be due to their species diversity and abundance (larger species group (Berendzen et al., 2008), and a recent study pool of potential colonizers), as well as direct development showing that parts of the Ozark and Ouachita fish faunas are and increased terrestriality (and thus improved dispersal distantly related due to a long period of drainage isolation abilities) relative to other plethodontids. For example, more (Hoagstrom et al., 2013). than 75% of the current range of the widely distributed One possible explanation for the rarity of plethodontid Plethodon cinereus was glaciated and uninhabitable just dispersal between the Ozarks and Ouachitas despite their 21,000 years ago (Highton & Webster, 1976), indicating a geographical proximity is that they were/are separated by a great capacity for dispersal in some species. relatively impermeable barrier. The Ozarks and Ouachitas Although habitats of the Ozarks and Ouachitas are differ- were formed by disparate geological processes in the Pennsyl- ent, habitats in the Ozarks and plateaus of the Eastern High- vanian (Thornbury, 1965), and have been disjunct highlands lands (Cumberland and Interior Low Plateaus) are similar, since their formation. Although the lower Arkansas River with karst topography and abundant springs. Regardless of may not have followed its present course bisecting the Inte- whether formerly continuous northern uplands linked the rior Highlands until the Pleistocene (Quinn, 1958; Galloway Ozarks and eastern plateaus prior to the Pleistocene as pos- et al., 2011), the primary drainages of the Interior Highlands ited by Mayden (1985), this karst topography is continuous have probably not changed appreciably since the Pennsylva- across southern Illinois and Indiana, linking the Ozarks and nian (Pfleiger, 1971). The Ozarks and Ouachitas have always eastern plateaus geologically. Several plethodontids are asso- been separated by lowland habitat (the Arkoma Basin; Carl- ciated with karst topography in both the Ozarks and Eastern ton & Cox, 1990), which would limit dispersal in many Highlands, particularly Eurycea lucifuga and E. longicauda. plethodontid salamanders. However, due to the narrow The distributions of karst-associated cave crayfishes (genus width of this putative barrier, other factors may have con- Orconectes; Crandall & Templeton, 1999) and fishes such as tributed to the infrequency of dispersal. Typhlichthys subterraneus (Niemiller et al., 2012), Forbe- The size and course of the Mississippi River have been fluid sichthys agassizi (Page & Burr, 2011), and Cottus carolinae over geological time, and it only reached its current northern (which exhibits recent gene flow via southern Illinois; extent in the middle Miocene (Galloway et al., 2011), approxi- Strange & Burr, 1997) spanning the Mississippi River also mately ~16–12 Ma. However, the Eastern and Interior High- seem to support habitat continuity from the eastern plateaus lands have been separated south of the present-day confluence to the Ozarks. Therefore, habitat similarity may help explain of the Mississippi and Ohio Rivers by a lowland barrier (the the frequency of dispersal and successful establishment from Mississippi Embayment) since the end of the Cretaceous the Eastern Highlands to the Ozarks. The Mississippi River’s (~65 Ma; Cushing et al., 1964). The Mississippi Embayment historically smaller size and lower northern extent may have and lower Mississippi River have limited dispersal by many also facilitated dispersal. terrestrial organisms (Soltis et al., 2006; Lemmon et al., 2007; Finally, the smaller number of plethodontid species in the Pyron & Burbrink, 2010). Therefore, the relative frequency of Interior Highlands has limited the pool of potential dis- plethodontid dispersal across this barrier is surprising, but sev- persers between the Ozarks and Ouachitas, and from the eral possible explanations exist. Interior Highlands to the Eastern Highlands. Our reconstruc- Ecological contingencies may have influenced dispersal; tions indicated that no Interior Highland plethodontids have the permeability of intervening lowland habitats may not be recolonized the Eastern Highlands. However, some fishes as important as the similarity (or disparity) of habitats in (Martin & Bonett, unpublished), crayfishes (Crandall & Tem- these adjacent uplands. The different geologies of the Ozarks pleton, 1999), and cryptobranchid salamanders (Sabatino & and Ouachitas have consequences for the types of habitats Routman, 2009) have probably dispersed from the Ozarks to available to plethodontid salamanders. The Ozarks are char- the Eastern Highlands. The permeability of lowland barriers, acterized by lower relief upland streams fed by abundant contingencies of habitat similarity or disparity and varying springs, due to the karst topography. This habitat is sizes of species pools for potential colonization are not amenable to species dependent on consistent water sources mutually exclusive hypotheses, and all may have contributed

418 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd Dispersal by salamanders among North American highland areas to the observed patterns of plethodontid dispersal and Pleistocene origins (Mayden, 1988; Strange & Burr, 1997; colonization. We note that dispersal may have been more Near & Keck, 2005; Berendzen et al., 2008). frequent, but its signature has been erased by extinction; upland habitats offer poor conditions for fossil formation ACKNOWLEDGEMENTS and salamanders’ fragile bones are rarely preserved. We thank K. Irwin, B. Thesing, K. Kozak and F. Burbrink for collecting some tissues or providing sequence data. K. Additional evidence for dispersal events Roberts and K. Irwin helped us review collection records Because our phylogenetic analysis did not include sequences for Desmognathus brimleyorum north of the Arkansas River. of cytochrome oxidase 1 (COI), the only geographically well- M. Treglia helped generate the relief map figure. Comments sampled marker for the widespread four-toed salamander from R. Bryson and four anonymous referees substantially (Hemidactylium scutatum), we were unable to reconstruct its improved the manuscript. Funding for this work was pro- phylogeographical history in concert with other plethodon- vided by National Science Foundation DEB1050322 (R.M.B.) tids. Herman (2009) concluded that the ancestors of the and DEB1210859 (R.M.B., M.A.S.), AMNH Theodore Roo- Ozark and Ouachita populations of H. scutatum indepen- sevelt Memorial Grant and ASIH Gaige Fund Award dently dispersed from the Eastern Highlands, 0.2 Ma and (M.A.S.), ODWC State Wildlife Grant (R.M.B.), and AGFC 0.4 Ma respectively. If this reconstruction is accurate, it State Wildlife Grant (D.B.S.). would add two more dispersal events from the Eastern High- lands to the Interior Highlands, and would represent the only Pleistocene plethodontid dispersal. REFERENCES Historical collection records exist for Desmognathus ‘brim- Allen, R.T. (1990) Insect endemism in the Interior Highlands leyorum’ from north of the Arkansas River in Arkansas [e.g. of North America. Florida Entomologist, 73, 539–569. – Crawford (KU 4601 4604), Pope, Pulaski and Faulkner Berendzen, P.B., Simons, A.M., Wood, R.M., Dowling, T.E. counties; K. Roberts, pers. comm.]. Most of these localities & Secor, C.L. (2008) Recovering cryptic diversity and are located in the Arkansas River Valley just south of the ancient drainage patterns in eastern North America: his- Ozark Plateau, but potentially represent one or more disper- torical biogeography of the Notropis rubellus species group sal events from the Ouachitas northward across the Arkansas (Teleostei: Cypriniformes). Molecular Phylogenetics and River. There were also formerly populations of Desmognathus Evolution, 46, 721–737. conanti on Crowley’s Ridge in north-eastern Arkansas (a nar- Blum, M.D., Guccione, M.J., Wysocki, D.A., Robnett, P.C. & row highland ridge that emerges from the Mississippi River Rutledge, E.M. (2000) Late Pleistocene evolution of the floodplain west of the river; Means, 1974; Trauth et al., lower Mississippi River valley, southern Missouri to – 2004). These populations represent an additional east west Arkansas. Geological Society of America Bulletin, 112, dispersal, potentially across the Mississippi River. Alterna- 221–235. tively, these populations may have become isolated on the Bonett, R.M. & Chippindale, P.T. (2004) Speciation, phylo- west side of the Mississippi River when the river changed geography and evolution of life history and morphology in from its former course west of Crowley’s Ridge to its current plethodontid salamanders of the Eurycea multiplicata com- course to the east (Blum et al., 2000). Unfortunately, plex. Molecular Ecology, 13, 1189–1203. Desmognathus have not been found on Crowley’s Ridge since Bonett, R.M., Steffen, M.A., Lambert, S.M., Wiens, J.J. & before 1990 (Trauth et al., 2004), and we cannot determine Chippindale, P.T. (2014) Evolution of paedomorphosis in the relationships or timing of colonization of these popula- plethodontid salamanders: ecological correlates and re-evo- tions without genetic samples. lution of metamorphosis. Evolution, 68, 466–482. Carlton, C.E. & Cox, R.T. (1990) A new species of Arianops KEY CONCLUSIONS from Central Arkansas and biogeographic implications of the Interior Highlands Arianops species (Coleoptera: Pse- Although the Ozark and Ouachita uplands are geographically laphidae). The Coleopterists’ Bulletin, 44, 365–371. proximate, they have had little faunal exchange, due to inter- Chippindale, P.T., Bonett, R.M., Baldwin, A.S. & Wiens, J.J. vening barriers, differing habitats and/or smaller species (2004) Phylogenetic evidence for a major reversal of life- pools. Instead, most plethodontids dispersed from the East- history evolution in plethodontid salamanders. Evolution, ern to the Interior Highlands. Furthermore, most of these 58, 2809–2822. dispersal events occurred during the Miocene and Pliocene, Crandall, K.A. & Buhay, J.E. (2008) Global diversity of cray- although better sampling and additional data are needed to fish (Astacidae, Cambaridae, and Parastacidae—Decapoda) resolve recent dispersal events (e.g. Hemidactylium scutatum, in freshwater. Hydrobiologia, 595, 295–301. Plethodon albagula). Taken together, our findings indicate Crandall, K.A. & Templeton, A.R. (1999) The zoogeography substantial but asymmetrical faunal exchange between the and centers of origin of the crayfish subgenus Procericam- Eastern and Interior Highlands, and support a growing barus (Decapoda: Cambaridae). Evolution, 53, 123–134. consensus that many Interior Highland species have pre-

Journal of Biogeography 43, 410–422 419 ª 2015 John Wiley & Sons Ltd S. D. Martin et al.

Cross, F.B., Mayden, R.L. & Stewart, J.D. (1986) Fishes in don). Proceedings of the Royal Society B: Biological Sciences, the western Mississippi basin (Missouri, Arkansas, and 273, 539–546. Red rivers). The zoogeography of North American freshwater Kozak, K.H., Mendyk, R.W. & Wiens, J.J. (2009) Can parallel fishes (ed. by C.H. Hocutt and E.O. Wiley), pp. 363–412. diversification occur in sympatry? Repeated patterns of John Wiley & Sons, New York, NY. body-size evolution in coexisting clades of North Ameri- Cushing, E.M., Boswell, E.H. & Hosman, R.L. (1964) General can salamanders. Evolution, 63, 1769–1784. geology of the Mississippi embayment. USGS Professional Lemmon, E.M., Lemmon, A.R. & Cannatella, D.C. (2007) Paper, 448-B,1–28. Geological and climatic forces driving speciation in the Darriba, D., Taboada, G.L., Doallo, R. & Posada, D. (2012) continentally distributed trilling chorus frogs (Pseudacris). jModelTest 2: more models, new heuristics and parallel Evolution, 61, 2086–2103. computing. Nature Methods, 9, 772. Lemmon, A.R., Brown, J.M., Stanger-Hall, K. & Lemmon, Dowling, H.G. (1956) Geographic relations of Ozarkian E.M. (2009) The effect of ambiguous data on phylogenetic amphibians and reptiles. The Southwestern Naturalist, 1, estimates obtained by maximum likelihood and Bayesian 174–189. inference. Systematic Biology, 58, 130–145. Drummond, A.J. & Rambaut, A. (2007) Tracer v1.4. Avail- MacArthur, R.H. & Wilson, E.O. (1963) An equilibrium the- able at: http://beast.bio.ed.ac.uk/Tracer ory of insular zoogeography. Evolution, 17, 373–387. Drummond, A.J., Suchard, M.A., Xie, D. & Rambaut, A. MacArthur, R.H. & Wilson, E.O. (1967) The theory of Island (2012) Bayesian phylogenetics with beauti and the beast biogeography. Princeton University Press, Princeton, NJ. 1.7. Molecular Biology and Evolution, 29, 1969–1973. Maddison, W.P. (1997) Gene trees in species trees. Systematic Duellman, W.E. & Sweet, S.S. (1999) Distribution patterns of Biology, 46, 523–536. amphibians in the Nearctic region of North America. Pat- Mayden, R.L. (1985) Biogeography of Ouachita highland terns of distribution of amphibians: a global perspective (ed. fishes. The Southwestern Naturalist, 30, 195–211. by W.E. Duellman), pp. 31–109. Johns Hopkins University Mayden, R.L. (1987) Pleistocene glaciation and historical Press, Baltimore, MD. biogeography of North American central highland fishes. Galloway, W.E., Whiteaker, T.L. & Ganey-Curry, P. (2011) Quaternary Environments of Kansas. Kansas Geological History of Cenozoic North American drainage basin evo- Survey Guidebook Series, 5, 141–151. lution, sediment yield, and accumulation in the Gulf of Mayden, R.L. (1988) Vicariance biogeography, parsimony, Mexico basin. Geosphere, 7, 938–973. and evolution in North American freshwater fishes. Sys- Guindon, S. & Gascuel, O. (2003) A simple, fast and accurate tematic Biology, 37, 329–355. method to estimate large phylogenies by maximum-likeli- McCormack, J.E., Huang, H. & Knowles, L.L. (2009) Sky hood. Systematic Biology, 52, 696–704. islands. Encyclopedia of Islands, 4, 841–843. Heald, W.F. (1951) Sky islands of Arizona. Natural History, Means, D.B. (1974) The status of Desmognathus brimleyorum 60,95–96. Stejneger and an analysis of the genus Desmognathus (Am- Herman, T.A. (2009) Range-wide phylogeography of the phibia, Urodela) in Florida. Bulletin of the Florida State four-toed salamander (Hemidactylium scutatum): out of Museum of Biological Sciences, 18,1–100. Appalachia and into the glacial aftermath. Unpubl. MS Miller, M.A., Pfeiffer, W. & Schwartz, T. (2010) Creating the thesis, Bowling Green State University, Bowling Green. CIPRES Science Gateway for inference of large phyloge- Highton, R. & Webster, T.P. (1976) Geographic protein vari- netic trees. Proceedings of the Gateway Computing Environ- ation and divergence in populations of the salamander ments Workshop (GCE), 14 November 2010, pp. 1–8. IEEE, Plethodon cinereus. Evolution, 30,33–45. New Orleans, LA. Hoagstrom, C.W., Ung, V. & Taylor, K. (2013) Miocene riv- Mueller, R.L. (2006) Evolutionary rates, divergence dates, ers and taxon cycles clarify the comparative biogeography and the performance of mitochondrial genes in Bayesian of North American highland fishes. Journal of Biogeogra- phylogenetic analysis. Systematic Biology, 55, 289–300. phy, 41, 644–658. Mueller, R.L., Macey, J.R., Jaekel, M., Wake, D.B. & Boore, Jiang, W., Chen, S.Y., Wang, H., Li, D.Z. & Wiens, J.J. J.L. (2004) Morphological homoplasy, life history evolution, (2014) Should genes with missing data be excluded from and historical biogeography of plethodontid salamanders phylogenetic analyses? Molecular Phylogenetics and Evolu- inferred from complete mitochondrial genomes. Proceedings tion, 80, 308–318. of the National Academy of Sciences USA, 101, 13820–13825. Kozak, K.H. & Wiens, J.J. (2012) Phylogeny, ecology, and Near, T.J. & Keck, B.P. (2005) Dispersal, vicariance, and tim- the origins of climate–richness relationships. Ecology, 93, ing of diversification in Nothonotus darters. Molecular S167–S181. Ecology, 14, 3485–3496. Kozak, K.H., Weisrock, D.W. & Larson, A. (2006) Rapid lin- Niemiller, M.L., Near, T.J. & Fitzpatrick, B.M. (2012) Delim- eage accumulation in a non-adaptive radiation: phyloge- iting species using multilocus data: diagnosing cryptic netic analysis of diversification rates in eastern North diversity in the southern cavefish, Typhlichthys subterraneus American woodland salamanders (Plethodontidae: Pletho- (Teleostei: Amblyopsidae). Evolution, 66, 846–866.

420 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd Dispersal by salamanders among North American highland areas

Niemiller, M.L., Fitzpatrick, B.M., Shah, P., Schmitz, L. & ated with stream drainages in a terrestrial salamander, Near, T.J. (2013) Evidence for repeated loss of selective Plethodon caddoensis. Molecular Phylogenetics and Evolu- constraint in rhodopsin of amblyopsid cavefishes (Teleos- tion, 59, 399–411. tei: Amblyopsidae). Evolution, 67, 732–748. Soltis, D.E., Morris, A.B., McLachlan, J.S., Manos, P.S. & Ouachita Ecoregional Assessment Team, The Nature Conser- Soltis, P.S. (2006) Comparative phylogeography of ungla- vancy. (2003) Ouachita Mountains Ecoregional Conserva- ciated eastern North America. Molecular Ecology, 15, tion Assessment. The Nature Conservancy Midwestern 4261–4293. Resource Office, Minneapolis, MN. Strange, R.M. & Burr, B.M. (1997) Intraspecific phylo- Ozarks Ecoregional Assessment Team, The Nature Conser- geography of North American highland fishes: a test of vancy. (2003) Ozarks Ecoregional Conservation Assessment. the Pleistocene vicariance hypothesis. Evolution, 51, The Nature Conservancy Midwestern Resource Office, 885–897. Minneapolis, MN. Tamura, K., Dudley, J., Nei, M. & Kumar, S. (2007) MEGA: Page, L.M. & Burr, B.M. (2011) A field guide to freshwater Molecular Evolutionary Genetics Analysis (MEGA) soft- fishes: North America north of Mexico. Houghton Mifflin ware. Molecular Biology and Evolution, 24, 1596–1599. Harcourt, Boston, MA. Thornbury, W.D. (1965) Regional geomorphology of the Pelletier, T.A., Crisafulli, C., Wagner, S., Zellmer, A.J. & Car- United States. Wiley and Sons, New York, NY. stens, B.C. (2014) Historical species distribution models Timpe, E.K. (2009) Exploring the faunal connection between predict species limits in western Plethodon salamanders. the Appalachian Mountains and the Ozark Plateau: a phy- Systematic Biology, in press. logeographic study of the long-tailed salamander complex Pfleiger, W.L. (1971) A distributional study of Missouri (Plethodontidae: Splerpinae: Eurycea). Unpubl. MS thesis, fishes. University of Kansas Publications, Museum of Natu- The University of Tulsa, Tulsa, OK. ral History, 20, 225–570. Trauth, S.E., Robison, H.W. & Plummer, M.V. (2004) The Pyron, R.A. & Burbrink, F.T. (2010) Hard and soft amphibians and reptiles of Arkansas. The University of allopatry: physically and ecologically mediated modes of Arkansas Press, Fayetteville, AR. geographic speciation. Journal of Biogeography, 37, Vieites, D.R., Min, M.S. & Wake, D.B. (2007) Rapid diversi- 2005–2015. fication and dispersal during periods of global warming by Pyron, R.A. & Wiens, J.J. (2011) A large-scale phylogeny of plethodontid salamanders. Proceedings of the National Amphibia including over 2800 species, and a revised classi- Academy of Sciences USA, 104, 19903–19907. fication of extant frogs, salamanders, and caecilians. Molec- Vieites, D.R., Roman, S.N., Wake, M.H. & Wake, D.B. ular Phylogenetics and Evolution, 61, 543–583. (2011) A multigenic perspective on phylogenetic rela- Quinn, J.H. (1958) Plateau surfaces of the Ozarks. Proceed- tionships in the largest family of salamanders, the ings of the Arkansas Academy of Science, 11,36–43. Plethodontidae. Molecular Phylogenetics and Evolution, Ree, R.H. & Smith, S.A. (2008) Maximum-likelihood 59, 623–635. inference of geographic range evolution by dispersal, Wiens, J.J. (2006) Missing data and the design of phyloge- local extinction, and cladogenesis. Systematic Biology, 57, netic analyses. Journal of Biomedical Informatics, 39,34– 4–14. 42. Roelants, K., Gower, D.J., Wilkinson, M., Loader, S.P., Biju, Wiens, J.J. & Morrill, M.C. (2011) Missing data in phyloge- S.D., Guillaume, K., Moriau, L. & Bossuyt, F. (2007) Glo- netic analysis: reconciling results from simulations and bal patterns of diversification in the history of modern empirical data. Systematic Biology, 60, 719–731. amphibians. Proceedings of the National Academy of Wiens, J.J., Engstrom, T.N. & Chippindale, P.T. (2006) Sciences USA, 104, 887–892. Rapid diversification, incomplete isolation, and the “speci- Sabatino, S.J. & Routman, E.J. (2009) Phylogeography and ation clock” in North American salamanders (genus conservation genetics of the hellbender salamander (Cryp- Plethodon): testing the hybrid swarm hypothesis of rapid tobranchus alleganiensis). Conservation Genetics, 10, 1235– radiation. Evolution, 60, 2585–2603. 1246. Wiley, E.O. & Mayden, R.L. (1985) Species and speciation in Shepard, D.B. & Burbrink, F.T. (2008) Lineage diversification phylogenetic systematics, with examples from the North and historical demography of a sky island salamander, American fish fauna. Annals of the Missouri Botanical Gar- Plethodon ouachitae, from the Interior Highlands. Molecu- den, 72, 596–635. lar Ecology, 17, 5315–5335. Zhang, P. & Wake, D.B. (2009) Higher-level salamander rela- Shepard, D.B. & Burbrink, F.T. (2009) Phylogeographic and tionships and divergence dates inferred from complete demographic effects of Pleistocene climatic fluctuations in mitochondrial genomes. Molecular Phylogenetics and Evolu- a montane salamander, Plethodon fourchensis. Molecular tion, 53, 492–508. Ecology, 18, 2243–2262. Zheng, Y. & Wiens, J.J. (2015) Do missing data influence the Shepard, D.B. & Burbrink, F.T. (2011) Local-scale environ- accuracy of divergence-time estimation with beast? Molec- mental variation generates highly divergent lineages associ- ular Phylogenetics and Evolution, 85,41–49.

Journal of Biogeography 43, 410–422 421 ª 2015 John Wiley & Sons Ltd S. D. Martin et al.

SUPPORTING INFORMATION BIOSKETCH

Additional Supporting Information may be found in the Samuel D. Martin primarily studied biogeography and online version of this article: macroevolution in minnows for his doctoral research. How- Appendix S1 GenBank accession numbers, ancestral area ever, his academic pursuits began with and still include coding, primer information, and species authorities. ecology and evolution, which is why his co-au- Appendix S2 Alternative ancestral area reconstruction, with thors haven’t ostracized him. transitions between the Eastern Highlands and western North Author contributions: R.M.B. and S.D.M. conceived the America allowed. study; all authors contributed sequence data and advised Appendix S3 Unpruned phylogram with BEAST support taxon sampling; S.D.M. analysed the data and prepared the values < 95% labelled. manuscript draft; S.D.M., R.M.B. and D.B.S. interpreted the results; and all authors read and approved the final manu- DATA ACCESSIBILITY script.

Genetic data used in this study can be found on GenBank; see Appendix S1 for a list of accession numbers. Editor: Robert Bryson Jr.

422 Journal of Biogeography 43, 410–422 ª 2015 John Wiley & Sons Ltd